🚀 Xinference v1.9.0 is here and it's HUGE!.We've got gpt-oss, Qwen-Image, and DeepSeek R1 with Function Calling! Rerank models now switch backends on the fly (vLLM included!), and sglang streams function calls. Time to upgrade your AI pipeline! ✨.#Xinference #AI.

Better virtualenv 💪with skip installed libs & individual model virtualenv toggle provides more flexibility.👍.Set environment variables when loading models for customized configurations. Improved virtualenv allows skipping system dependencies & customizing per model.🛠️.

Exciting Updates from Xinference v1.8.1!😊.Full support for GLM 4.5 series & Qwen3 models, experimental CUDA 12.8 image release (v1.8.1-cu128), 👏and automatic max supported length usage when max_tokens not set. #Xinference #Update 🚀.

Aligned with Enterprise Edition v0.1.2！ ✌️.🏢 Enhanced Prometheus metric monitoring, max_tokens now supports longest tokens by default. 🥳Resolving multimodal model compatibility issues for vLLM and Transformers inference, ensuring stable distributed deployments. #Update 🚀.

Exciting update in Xinference v1.8.0! 🚀Embedding support for llama.cpp backend, FLUX.1-Kontext-dev image editing model, and ERNIE and GLM-4.1V-Thinking from ERNIE Bot 4.5!.Expand cutting-edge model ecosystems and make thought of chain smarter.👏.

Xinference v1.7.1 is coming! 🥳.🙌Now support model deployment on Cambricon chips (MLU), introducing experimental traditional machine learning model inference!.We now enables vLLM v1 engine and MLX distributed inference.👍.Updated Helm support for vGPU environments with HAMi.🚀.

Discover the latest from Xinference v1.7.0! 🚀 . 🧠 Qwen3 model now supports Embedding & Reranker, with multi-engine switching in Embedding and advanced video 🎞️ generation—including first & last frame animation (just input start/end frames and let AI handle the rest!). #AI.

Exciting Enterprise Edition update! 🏢 .👏Enhance stability with master-slave synchronization for high availability. Enjoy improved features and interfaces, boosting multi-modal model experiences and system observability. 🙌.Elevate your business continuity! #Update 🚀.

Exciting update from Xinference v1.6.1! 🚀 .🥳Introducing Deepseek-R1-0528 & Qwen3 support, revamped Transformers VL model logic for continuous batching, and llama.cpp's Auto NGL for enhanced GPU deployment stability. ⌨️.Empowering AI capabilities! #Xinference #AI #Update 🧠🧩.

Xinference v1.6.0 is out! 🤩. Enjoy UI support for video & audio models🖥️, flexible Qwen3 inference control⚙️, stable environment isolation with Xoscar (PVLDB-accepted), and xllamacpp engine for llama.cpp. 🥳. Dive into multimodal AI with ease! 🙌.#AI.

🥳Exciting news! Xinference v1.5.1 is here!🙌.🧠Dive into Qwen3/Qwen3-MOE large model and explore Wan 2.1 text-to-video modeling for cutting-edge.Al video creation!👈.💡Experience VLLM's GGUFV2 format support with.stronger capabilities! #Xinference #Al.

Updates for Xinference's Enterprise Edition! .🏢 Unlock the Text to Video module interface for enhanced AI video creation. 🎥 Experience expanded Ascend Adaptation Capability, ensuring stable and efficient model performance on Ascend. #Xinference #EnterpriseEdition.

Exciting feature upgrades in Xinference! 👏.Gradio chat now showcases thinking process, Vision model offers input resolution control, and model downloads include progress tracking. Default settings are optimized for efficiency, with InternVL3 supporting AWQ inference. 🤩.

Exciting news! Xinference v1.5.0 is back with a bang! 🚀 .Introducing the Model Virtual Space feature, ensuring model stability by isolating dependencies. Embrace seamless model operations! .🌟 Check out the documentation for setup details:

🔥 Exciting news! Xinference v1.4.1 is here! 🎉 .Dive into vLLM distributed reasoning - run vLLM across multiple machines for efficient inference. SGLang engine now supports visual models. 👀.Plus, enjoy faster GPTQ quantized inference speed on the Transformer engine! 🤩 #LLM.

📢 Exciting news! Xinference v1.4.0 is out 🚀! .Featuring the return of Gemma-3 model and DeepSeek-v3 with Function Calling support. Enterprise Edition K8s now supports deploying clusters using Helm Charts ⛵. 🚀.Join us for more code contributions 💪💖. #Xinference 🌟.

🚀 Exciting news! Xinference v1.3.1 is here! 🎉 .🧐Adding support for the new Qwen model QwQ and Xllamacpp with continuous batching. 👐Plus, Qwen2.5-VL supports AWQ quantization for better efficiency! .➡️Check out the new reasoning_content parameter for enhanced parsing! #AI.

🚀 Big news! #Xinference v1.3.0.post1 is out! .Now, distribute DeepSeek V3/R1 across multiple machines—no Enterprise edition needed! .🖥️✨ Enhanced UI, 1-click DeepSeek, & new tools make it a breeze. Next up: vLLM & MLX for even better performance! #AI #TechUpdate.

🚀 Xinference v1.2.2 is here! .Deepseek r1 distillation model support is complete ✅. Get ready for deepseek v3 and R1 models in the upcoming 1.3.0 release 🎉. Enterprise Edition now features PD splitting support and improved UI! #AI #updates 👏 ⚡.

🐞 Bug fixes resolve compatibility issues with the OpenAI API and improve internationalization for a smoother experience! . 🌏 Enterprise version update optimizes support for Hiascend, Enflame, and Hygon platforms for better reliability! #Xinference #updates.